Hadoop MapReduce HA搭建

蔚落 2022-12-24 14:57 118阅读 0赞

MRAppMaster中如果ResourceManager 宕机了怎么办，与hadoop中的namenode问题一样，同样可以将  
RM搭建高可用。

###  ###

*  1、规划
 *  2、配置文件
 *   *  2.1 mapred-site.xml
     *  2.2 yarn-site.xml
     *  2.3 把配置文件传到 其他服务器上
     *  2.4 修改启动文件
 *  3、启动集群
 *  4、验证
 *   *  4.1 页面验证
     *  4.2 高可用验证

# 1、规划 #

准备四台虚拟机，规划如下

<table> 
 <thead> 
  <tr> 
   <th>主机</th> 
   <th>node2</th> 
   <th>node3</th> 
   <th>node4</th> 
   <th>node5</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>节点</td> 
   <td>NameNode</td> 
   <td>NameNode</td> 
   <td></td> 
   <td></td> 
  </tr> 
  <tr> 
   <td>节点</td> 
   <td></td> 
   <td>DataNode</td> 
   <td>DataNode</td> 
   <td>DataNode</td> 
  </tr> 
  <tr> 
   <td>节点</td> 
   <td></td> 
   <td>ZooKeper</td> 
   <td>ZooKeper</td> 
   <td>ZooKeper</td> 
  </tr> 
  <tr> 
   <td>节点</td> 
   <td>ZKFC</td> 
   <td>ZKFC</td> 
   <td></td> 
   <td></td> 
  </tr> 
  <tr> 
   <td>节点</td> 
   <td>JournalNode</td> 
   <td>JournalNode</td> 
   <td>JournalNode</td> 
   <td></td> 
  </tr> 
  <tr> 
   <td>节点</td> 
   <td></td> 
   <td></td> 
   <td>Resource Manager1</td> 
   <td>Resource Manager2</td> 
  </tr> 
  <tr> 
   <td>节点</td> 
   <td></td> 
   <td>Node Manager1</td> 
   <td>Node Manager2</td> 
   <td>Node Manager3</td> 
  </tr> 
 </tbody> 
</table>

在[hadoop基于zookeper自动高可用搭建][hadoop_zookeper]  
中已经搭建好了namdenode的高可用，所以本实例在其基础上再搭建即可，可以看到只需要在配置  
Resource Manager1，Resource Manager2即可，Node Manager1无需配置，其配置同DataNode配置，在slaves配置文件中。

# 2、配置文件 #

## 2.1 mapred-site.xml ##

进入到hadoop配置文件目录下 生成mapred-site.xml文件

cd $HADOOP_HOME/etc/hadoop
    cp mapred-site.xml.template  mapred-site.xml
    vim  mapred-site.xml

增加如下配置

<property>
       <name>mapreduce.framework.name</name>
       <value>yarn</value>
    </property>

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70]

## 2.2 yarn-site.xml ##

在进入到hadoop配置文件目录下 ,yarn-site.xml中添加如下内容

<property>
       <name>yarn.nodemanager.aux-services</name>
       <value>mapreduce_shuffle</value>
    </property>
    
    <property>
       <name>yarn.resourcemanager.ha.enabled</name>
       <value>true</value>
     </property>
    
     <property>
       <name>yarn.resourcemanager.cluster-id</name>
       <value>cluster1</value>
     </property>
    
     <property>
       <name>yarn.resourcemanager.ha.rm-ids</name>
       <value>rm1,rm2</value>
     </property>
    
     <property>
       <name>yarn.resourcemanager.hostname.rm1</name>
       <value>node4</value>
     </property>
    
     <property>
       <name>yarn.resourcemanager.hostname.rm2</name>
       <value>node5</value>
     </property>
    
     <property>
       <name>yarn.resourcemanager.zk-address</name>
       <value>node3:2181,node4:2181,node5:2181</value>
     </property>

## 2.3 把配置文件传到 其他服务器上 ##

本实例在node2上修改了 yarn-site.xml， mapred-site.xml，将hadoop所有配置文件都传到其他服务器上，以达到共享

scp * node3:`pwd` && scp * node4:`pwd` && scp * node5:`pwd`

## 2.4 修改启动文件 ##

先启动yarn 再启动rm

start-yarn.sh
     yarn-daemon.sh start resourcemanager

新增hdfs集群 start\_hdfs\_ha.sh启动文件,在home目录下

vim   start_hdfs_ha.sh

添加如下内容，实现一键启动

#!/bin/bash
    for node in node3 node4 node5
    do
      ssh $node "source /etc/profile; zkServer.sh start"
    done
    
    sleep 1
    start-dfs.sh
    
    ssh node4 "source /etc/profile; start-yarn.sh"
    ssh node5 "source /etc/profile; yarn-daemon.sh start resourcemanager"
    
    echo "-----------node2-jps----------- "
    jps
    for node in node3 node4 node5
    do
    echo "-----------$node-jps----------- "
     ssh $node "source /etc/profile; jps"
    done

再添加一键停服务脚本

#!/bin/bash
    ssh node4 ". /etc/profile; stop-yarn.sh"
    ssh node3 ". /etc/profile; yarn-daemon.sh stop resourcemanager"
    
    stop-dfs.sh
    sleep 1
    
    for node in node3 node4 node5
    
    do
      ssh $node "source /etc/profile; zkServer.sh stop"
    done
    
    echo "-----------node2-jps----------- "
    jps
    
    for node in node3 node4 node5
    do
    echo "-----------$node-jps----------- "
     ssh $node "source /etc/profile; jps"
    
    done

将启动和命令 传到其他服务器node\[3-5\]上

scp start_hdfs_ha.sh  node3:`pwd` && scp start_hdfs_ha.sh  node4:`pwd` && scp start_hdfs_ha.sh  node5:`pwd`
    scp stop_hdfs_ha.sh  node3:`pwd` && scp stop_hdfs_ha.sh  node4:`pwd` && scp stop_hdfs_ha.sh  node5:`pwd`

# 3、启动集群 #

任意一台服务器执行

./start_hdfs_ha.sh  
    启动后可以看到 node4 和node5上有resourcemanager进程以及nodemanager进程

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 1]

# 4、验证 #

## 4.1 页面验证 ##

RM的默认端口为8088，浏览器输入node4:8088.看到node4的resourcemanager处于active状态  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 2]  
http://node5:8088/  
![在这里插入图片描述][20201129003600371.png]  
输入node5:8088后出现上述页面后会自动调转到node4:8088，是由于node5处于standby状态。  
http://node5:8088/

## 4.2 高可用验证 ##

在node4上看到rm的进程为 1842  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 3]  
kill 掉rm进程模拟宕机

kill -9 1842
    jps

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 4]  
然后  
再刷新node5:8088

看到node5上的rm自动切换为active  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 5]  
然后在启动node4上的rm。

yarn-daemon.sh start resourcemanager
    jps

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 6]  
启动后，再刷新node4:8088 看到 node4启动rm后rm为standby状态

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 7][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 8][]

[hadoop_zookeper]: https://blog.csdn.net/whandgdh/article/details/109710266
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70]: /images/20221120/8a2f7da78b4b44dfb31952eaaa2b9155.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/20201129003152758.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20201129003525631.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[20201129003600371.png]: https://img-blog.csdnimg.cn/20201129003600371.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 3]: https://img-blog.csdnimg.cn/20201129004030361.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 4]: https://img-blog.csdnimg.cn/20201129004123834.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 5]: https://img-blog.csdnimg.cn/20201129004228612.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 6]: https://img-blog.csdnimg.cn/20201129004449842.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 7]: https://img-blog.csdnimg.cn/20201129004531249.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro_size_16_color_FFFFFF_t_70 8]: https://img-blog.csdnimg.cn/20201129004511129.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3doYW5kZ2Ro,size_16,color_FFFFFF,t_70